PDF Data Extraction & Accuracy Analysis

This repository contains a set of Python tools developed for extracting data from scanned PDF documents and measuring the accuracy of compiled datasets against original source files.

The project was created while working with a collection of hospital annual reports. Most of the source files were image-based scans, which required OCR processing before any meaningful comparison could be performed.

---

Overview

The tools support the following workflow:

1. Extract text from standard (text-based) PDFs
2. Perform OCR on scanned PDF pages
3. Convert extracted content into Word and Excel formats
4. Compare a compiled Excel/PDF dataset against the original 15 source reports
5. Calculate match rate and overall data accuracy
6. Generate detailed difference reports

---

Scripts

Script	Description
extract_simple.py	Extracts text from PDFs that contain selectable text and exports results to Word and Excel
extract_ocr.py	Converts scanned PDF pages into images and applies OCR
compare_excel_vs_15pdfs.py	Performs OCR on multiple reports and compares extracted numbers against a compiled file
data_difference_checker.py	Compares character counts, word counts, and unique numbers between an original text source and a processed file
full_15pdf_accuracy_checker.py	Starts from the original 15 PDF files, extracts all unique numbers, and calculates match rate against the target Excel/PDF

---

Requirements

pip install -r requirements.txt

System dependencies:

• Tesseract OCR — Windows installer
• Poppler — required by pdf2image for converting PDF pages to images

After installation, update the following paths inside the relevant scripts:

TESSERACT_PATH = r"C:\Program Files\Tesseract-OCR\tesseract.exe"
POPPLER_PATH   = r"C:\poppler\Library\bin"

Also update the input folder and file paths to match your local directory structure.

---

Usage

# Basic text extraction
python extract_simple.py

# OCR extraction from scanned PDFs
python extract_ocr.py

# Compare compiled file against previously extracted text
python data_difference_checker.py

# Full accuracy check starting from the original 15 PDFs
python full_15pdf_accuracy_checker.py

---

Accuracy Calculation

The primary accuracy metric is based on unique numerical values:

Accuracy = (Matched Numbers / Total Unique Numbers in Source) × 100

Reports include:

• Total unique numbers found in the source files
• Numbers successfully matched in the target file
• Numbers missing from the target file
• Numbers present only in the target file
• Overall match rate / accuracy percentage

---

Output

Each analysis script generates:

• A detailed text report
• An Excel summary with matched, missing, and extra values
• Clear accuracy percentage for quick review

---

Notes

• OCR quality depends heavily on the resolution and clarity of the original scans.
• The first run of any OCR-based script can take a significant amount of time.
• Intermediate OCR text files are saved locally so subsequent runs are much faster.
• These scripts were written for a specific set of annual reports and will require path updates for use with other documents.

---

Future Improvements

• Image preprocessing (deskew, contrast enhancement) before OCR
• More robust table structure detection
• Configurable paths through an external settings file
• Per-document accuracy breakdown